Analysis of error profiles of indels and structural variants in deep sequencing data

In this work, we generated ultra-deep sequencing data using our previously established dilution models (COLO829) on known somatic indels (n=23) and SVs (n=17). We discovered that the error rate of indels and SVs are 100- to >1000-fold lower than that of SNVs. This finding was fully recapitulated in our analysis of 309 indels and 1153 SVs discovered from a relapsed B-ALL cohort of 103 patients, although homopolymer indels can have high error rates (>1%). Our data also indicated that the number of repeating units are highly predictive relative to the error rate of homopolymer indels. Next, we assayed end-of-induction remission samples from 72 B-cell lymphoblastic leukemia patients that relapsed by selecting ~5 somatic clonal SNV/Indel/SV markers, which confirmed that SVs and indels have >10-fold lower error rates than SNVs. Our next generation sequencing (NGS) approach had 44 positive detections and outperformed the current standard method of clinical flow cytometry (n=37; 51%) for detecting minimal residual disease. The NGS-based method detected 92% of designed markers for samples with MRD >0.3%, and this detection rate dropped to 27% for MRD between 0.1% and 0.01%, indicating the difficulty in recovering mutant molecules when their frequencies are very low.